Data Science Practice

---

---

---
exclude: true

---

## Plan for this week
 
 * **[L04] Data Visualisation with `{ggplot2}`**
 
 + Motivation
 + `{ggplot2}` package
 
 + Aesthetic attributes
 + Geometric objects
 + Facets
 
 * **[L05] Examples**
 
 * **[L06] General Comments**
 
 + Which (common) plot to use?
 + What to pay attention to?
 + Further comments

---
class: inverse, center, middle

# Motivation

---

## Why graphs are important?

* *First impression* matters -- it is visually stimulating.
 
* *Efficiency in exploring the data* -- Always visualise your data sets before creating any models!
 
* *Effective communication* -- 'A picture is worth a thousand words'.
 
* Sometimes, summary statistics are just not enough -- see examples [here](https://www.autodeskresearch.com/publications/samestats).
 
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-1-1.png" width="100%" style="display: block; margin: auto;" />
]

---

## Why graphs are important?

.pull-left[
 
* *First impression* matters -- it is visually stimulating.
 
* *Efficiency in exploring the data* -- Always visualise your data sets before creating any models!
 
* *Effective communication* -- 'A picture is worth a thousand words'.
 
* Sometimes, summary statistics are just not enough -- see examples [here](https://www.autodeskresearch.com/publications/samestats).
 
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-1A-1.png" width="100%" style="display: block; margin: auto;" />
]

---

## ggplot

The name `ggplot` comes from the book *The Grammar of Graphics* by Leland Wilkinson (2005) (ref: ISBN 978-0-387-98774-3). A grammar of graphics is a framework that allows a structured and layered approach to construct graphics.

#### Components of a graph

<img src="data:image/png;base64,#./img/components_of_ggplot.jpg" width="65%" style="display: block; margin: auto;" />
[Image source](https://towardsdatascience.com/a-comprehensive-guide-to-the-grammar-of-graphics-for-effective-visualization-of-multi-dimensional-1f92b4ed4149)

---

---

## The `{ggplot2}` package
An R visualisation package developed by Hadley Wickham (2007) that adapts and implements the concept of *ggplot*.

> 'ggplot2 (Wickham 2009) builds on Wilkinson’s grammar by focussing on the primacy of layers and adapting it for use in R. In brief, the grammar tells us that a graphic maps the data to the **aesthetic attributes** (colour, shape, size) of **geometric objects** (points, lines, bars). The plot may also include **statistical transformations** of the data and information about the plot’s **coordinate system**. **Facetting** can be used to plot for different subsets of the data. The combination of these independent components are what make up a graphic.' -- Hadley Wickham, [*ggplot2*](https://ggplot2-book.org/)

Check out the 'R Graph Gallery' [here](https://www.r-graph-gallery.com/).

---
## Aesthetic attributes

What attributes will be mapped onto the x-axis and y-axis?

]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-p0-1.png" width="100%" style="display: block; margin: auto;" />
]

---

## Scatter plot

Now we can actually add some points

```r
airquality %>% 
 ggplot(aes(x = Solar.R, y = Ozone)) +
* geom_point()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-p1-1.png" width="100%" style="display: block; margin: auto;" />
]

---
## Colours

How about adding the colour to points based on another (factor) variable.

```r
airquality %>% 
 ggplot(aes(x = Solar.R, y = Ozone, 
* color = factor(Month))) +
 geom_point()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-p3-1.png" width="100%" style="display: block; margin: auto;" />
]

---

## Geometric objects

We have seen points, what else?

```r
airquality %>% 
 ggplot(aes(x = Solar.R, y = Ozone)) + 
 geom_point() + 
* geom_smooth()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-p4-1.png" width="100%" style="display: block; margin: auto;" />
]

---
## Geometric objects

```r
airquality %>% 
 ggplot(aes(x = Solar.R, y = Ozone)) + 
* geom_hex()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-p5-1.png" width="100%" style="display: block; margin: auto;" />
]

---
## Geometric objects

```r
airquality %>% 
 ggplot(aes(x = factor(Month), 
 y = Ozone)) +
* geom_boxplot()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-p6-1.png" width="100%" style="display: block; margin: auto;" />

]

---

## Facets

Dividing data sets into sub groups and plot separately for each group. It is useful when when the relationship is beyond 2D -- you can explore relationship between two variables conditioned on other variable(s). There are two common types of faceting in R, `facet_grid` and `facet_wrap`.

```r
airquality %>% 
 ggplot(aes(x = Solar.R, y = Ozone)) + 
 geom_point() + 
* facet_wrap(~Month, nrow = 2)
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-p7-1.png" width="75%" style="display: block; margin: auto;" />
]

---

## Incorporating data processing

```r
airquality %>% na.omit() %>% 
 mutate(TempGp = cut(Temp, 
 breaks = quantile(Temp, (0:4)/4), 
* nc = TRUE)) %>% # use %>%
 mutate(WindGp = cut(Wind, 
 breaks = quantile(Wind, (0:4)/4), 
 inc = TRUE)) %>% 
 ggplot(aes(x = Solar.R, y = Ozone)) + 
* geom_point() + # use +
 facet_grid(WindGp ~ TempGp)
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L01-p8-1.png" width="100%" style="display: block; margin: auto;" />
]

---

---

# The NZ Vehicle Registration Data

---
## Car registration open data

[NZ vehicle registration open data](https://nzta.govt.nz/resources/new-zealand-motor-vehicle-register-statistics/new-zealand-vehicle-fleet-open-data-sets/) provides a snapshot of the currently registered fleets in NZ. The 2019 dataset has 206,099 rows with 34 columns.

```r
cars.df <- read_csv("datasets/VehicleYear-2019.csv")
# glimpse(cars.df)
dim(cars.df) # dimension of the data frame
head(names(cars.df), 20) # some column names of the cars.df
```

```
## [1] 206099     34
```

```
##  [1] "ALTERNATIVE_MOTIVE_POWER"    "BASIC_COLOUR"               
##  [3] "BODY_TYPE"                   "CC_RATING"                  
##  [5] "CHASSIS7"                    "CLASS"                      
##  [7] "ENGINE_NUMBER"               "FIRST_NZ_REGISTRATION_YEAR" 
##  [9] "FIRST_NZ_REGISTRATION_MONTH" "GROSS_VEHICLE_MASS"         
## [11] "HEIGHT"                      "IMPORT_STATUS"              
## [13] "INDUSTRY_CLASS"              "INDUSTRY_MODEL_CODE"        
## [15] "MAKE"                        "MODEL"                      
## [17] "MOTIVE_POWER"                "MVMA_MODEL_CODE"            
## [19] "NUMBER_OF_AXLES"             "NUMBER_OF_SEATS"
```

---

## Distribution of car weight

```r
cars.df %>%
 ggplot(aes(x = GROSS_VEHICLE_MASS)) +
 geom_histogram()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L02-p11-1.png" width="100%" style="display: block; margin: auto;" />

]

---
## Filter zero weight cars

```r
cars.df %>%
* filter(GROSS_VEHICLE_MASS > 0,
* POWER_RATING > 0) %>%
 ggplot(aes(x = POWER_RATING,
 y = GROSS_VEHICLE_MASS)) +
 geom_point()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L02-p12-1.png" width="100%" style="display: block; margin: auto;" />
]

---
## Better scatter plot

```r
cars.df %>%
 filter(GROSS_VEHICLE_MASS > 0,
 POWER_RATING > 0) %>%
 ggplot(aes(x = POWER_RATING,
 y = GROSS_VEHICLE_MASS)) +
* geom_point(alpha = 0.05)
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L02-p13-1.png" width="100%" style="display: block; margin: auto;" />
]

---

## Try hex(bin) plot

```r
cars.df %>%
 filter(GROSS_VEHICLE_MASS > 0,
 POWER_RATING > 0) %>%
 ggplot(aes(x = POWER_RATING,
 y = GROSS_VEHICLE_MASS)) +
* geom_hex()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L02-p14-1.png" width="100%" style="display: block; margin: auto;" />
]

---
## Further exploration

```r
cars.df %>%
 filter(GROSS_VEHICLE_MASS > 0,
 POWER_RATING > 0,
 MAKE == 'TOYOTA') %>%
* ggplot(aes(x = FIRST_NZ_REGISTRATION_YEAR,
* y = GROSS_VEHICLE_MASS)) +
 geom_point()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L02-p15-1.png" width="100%" style="display: block; margin: auto;" />

]

---

## Add 'jitter'

```r
cars.df %>%
 filter(GROSS_VEHICLE_MASS > 0,
 POWER_RATING > 0,
 MAKE == 'TOYOTA') %>%
 ggplot(aes(x = FIRST_NZ_REGISTRATION_YEAR,
 y = GROSS_VEHICLE_MASS)) +
* geom_jitter()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L02-p16-1.png" width="100%" style="display: block; margin: auto;" />
]

---

## Scales, labels and theme

If you are making plots for others, it is a good idea to make them clear and readable.

```r
p <- cars.df %>%
 filter(GROSS_VEHICLE_MASS > 0,
 POWER_RATING > 0) %>%
 ggplot(aes(x = POWER_RATING,
 y = GROSS_VEHICLE_MASS)) +
 geom_point(alpha = 0.05) +
* labs(title = "Engine power vs. Car weight",
* x = "Power rating (kw)",
* y = 'Vehicle mass (kg)',
* caption = 'Data from nzta.govt.nz')
p
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L02-p17-1.png" width="100%" style="display: block; margin: auto;" />
]

---

## Get the scale and coordinate right, and a different theme

```r
p +
* scale_y_continuous(
 limits = c(0,3500),
 breaks = seq(0,3500,by=500)) +
 coord_cartesian(xlim = c(0, 500)) +
* theme_minimal()
```
]
.pull-right[
<img src="data:image/png;base64,#img/W02L02-p18-1.png" width="100%" style="display: block; margin: auto;" />
]

---

---
## Which (common) plot to use?

#### Single Variable (univariate)

Type|(Common) Plot to use|Features to Pay Attention to
----|--------------------|----------------------------
Quantitative (e.g. a variable of measurement)| dot plot/stript chart, histogram, density plot, box plot|shape, peaks, center, variability, outliers.
Qualitative (e.g. count of a grouping variable)|bar plot, pie chart, table of counts|majority/minority group, gaps in group counts.

---

## Which (common) plot to use?

#### Two Variables (bivariate)

Type|(Common) Plot to use|Features to Pay Attention to
----|--------------------|----------------------------
Quantitative vs. Qualitative|side-by-side histogram/density/box plot|compare shapes, centers, variability; outliers from individual group.
Quantitative vs. Quantitative|scatter plot, line plot|shapes, peaks, center, variability, outliers, correlation, grouping of observations, seasonal variation (for time series).
Qualitative vs. Qualitative|faceted bar plot, 2-way table of counts, pie chart (?)|compare group counts, distributions and gaps.

NB: Plotting for 2+ variables can often be achieved by 'reducing' it to some variations of bi-variate plots.

---

## General Plotting Advice

* Use colors, shapes etc, but keep things **balanced**.
  
  * Keep the focus -- produce a plot with clear message in mind.
  
  * Be aware of scales, labels and Hierarchy.
  
  * Leave some white space. 
  
  * ...

---

## General Plotting Advice

* Avoid pie charts (?!)
  
     > "Avoid pie-charts. Especially 3d pie-charts. Especially 3d pie-charts with exploding wedges. I promise all my students an instant fail if I ever see anything so appalling." - Rob J Hyndman, from ["Twenty rules for good graphics"](https://robjhyndman.com/hyndsight/graphics/)

* Sometimes, it would be helpful to produce multiple (types of) plots for the same data to reveal the real pattern. 
  
     For example...

---

...For example, what can you see from the boxplot below?

---

But if we look at the density plot...

---
## Charts and accessibility

While charts are very much a visual medium, we can improve accessibility of our charts by including 'alternative text', often known as 'alt text'.

### 📚 To read

Cesal., A. (2020). *Writing Alt Text for Data Visualization*. Nightingale https://nightingaledvs.com/writing-alt-text-for-data-visualization/